Analytics
Train/Test Split
Sends a random share of the rows out of the Test pin and the rest out of Train, reproducibly from a seed, and labels each row in a Split column.
When to use it
Use it before fitting a model: fit on the Train rows, then check the model on Test rows it has not seen.
In a pipeline
File InputTrain/Test SplitLogistic RegressionScore Model
Settings
| Setting | Type | Required | Default or hint |
|---|---|---|---|
| Test share (0–1) | number | — | 0.2 |
| Seed | number | — | 1 |
| Keep the proportions of | select | — | Optional: a class column |
| Split column name | text | — | Default: Split |
Example
Hold out 40%
- Test share (0–1)=0.4
- Seed=1
| Id |
|---|
| 1 |
| 2 |
| 3 |
| 4 |
| 5 |
| Id | Split |
|---|---|
| 1 | Train |
| 2 | Test |
| 3 | Train |
| 4 | Test |
| 5 | Train |
Worth knowing
- Exactly round(rows × share) rows go to Test; the same seed always picks the same rows. A connection without a pin label takes Train.
- Keep the proportions of draws the share within each value of a class column, so a rare class appears on both pins.
- Want every row in one table? Connect both pins to Union Rows — the Split column says which part each row is in.